Increasingly, we are witnessing a growing number of developments in the field of robotics characterized by their intent to integrate man and machine in a safe and functional manner. The suitability of a manipulator to work in close proximity with humans is determined first by the level of safety it can guarantee towards its human counterparts. Guaranteeing safety is a difficult if not impossible exercise as we can rarely guarantee the dependability of the numerous components required to complete a modern manipulator. Add in the human factor, and our task becomes insurmountable.
Thus, much focus has been centered on interactive robots which are expected to perform in a safe and dependable manner in unknown and unpredictable environments. Arguably, the chief safety concern is the manipulator's response to collisions with humans. Such collisions are responsible for numerous injuries each year, despite the existence of barriers, and other fail-safe mechanisms. As we move closer and closer towards a shared environment between robots and humans, new approaches to manipulator design are becoming increasingly important.
Devices utilizing the unique properties of Magneto-Rheological (MR) and Electro-Rheological (ER) fluids have been developed for robotic applications, however almost entirely for use in haptic systems. While it has been suggested in the literature how such devices might be used in a manipulator to improve both safety and performance, there appears to be a general reluctance towards adopting such technology as a viable alternative to the current solutions.
There are several sources of danger when working closely with robotic devices. However, collisions involving robots and humans pose arguably the largest degree of danger. It is expected that collisions will become unavoidable, if not routine as we continue to integrate man and machine into a single working environment.
In the prior art, friction and other nonlinearities present in the transmission of industrial manipulators led to the development of joint torque controlled systems. Characterized by torque sensors located at the joints, such manipulators are capable of achieving precise force control. Moreover, this class of manipulators can be controlled to exhibit very low impedance when operating within the controllable bandwidth. Barring the potential for high velocity collisions, i.e., collisions having their dynamics above the controllable frequency, this implementation can successfully attenuate the impact loads associated with collisions.
However, collisions occurring above the controllable bandwidth are subject to the open loop characteristics of the manipulator. As the demands for performance are increased, such systems require faster and more powerful actuators in order to successfully control and minimize impact forces during a collision. Any degree of safety introduced by a joint torque controlled system relies on the proper operation of system electronics. Thus guaranteeing collision safety becomes difficult as electronic components are susceptible to failure.
In attempts to guarantee the safety of humans within a shared workspace, much research has been focused on the development of manipulators which are intrinsically safe. That is, manipulators which by means of their mechanical properties can guarantee some level of collision safety in the absence of a controller.
To understand the degree of safety one might associate with a manipulator, one can examine the results of an uncontrolled collision through the use of the Head Injury Criterion (HIC). The HIC along with its variations have long been used by the automotive industry to gauge the severity of collisions. In the field of robotics, it can also be used to gain similar insight. The HIC is defined as:
                    HIC        =                              max                                          t                1                            ,                              t                2                                              ⁢                      {                                                            (                                                            t                      2                                        -                                          t                      1                                                        )                                ⁡                                  [                                                            1                                                                        t                          2                                                -                                                  t                          2                                                                                      ⁢                                                                  ∫                                                  t                          1                                                                          t                          2                                                                    ⁢                                                                        a                          ⁡                                                      (                            t                            )                                                                          ⁢                                                                                                  ⁢                                                  ⅆ                          t                                                                                                      ]                                            2.5                        }                                              (        1        )            where a is the acceleration of the head (in g's), and t1 and t2 are times within the collision selected to maximize the HIC, such that t1<t2.
An HIC of 100 is the maximum value considered to be non-life threatening. To gauge how the effective inertia of a link is related to a manipulator's inherent ability to collide safely, a single axis robot colliding with a human head is simulated. With reference to FIG. 1, the results of the HIC show that a manipulator's safety can be improved by reducing its effective inertia. This find inspired the generation of light weight robots.
One of the first manipulators to be designed under the lightweight paradigm was the Whole Arm Manipulator (WAM). The WAM uses steel cable transmission allowing actuators to be located at the manipulator's base. Removing the actuators from the links reduces the associated link mass, and hence the associated link inertia. This works to improve the inherent safety of the manipulator.
According to the prior art, one manipulator uses light weight carbon composites to form its links. Furthermore, an advanced actuator design integrated with low weight harmonic reduction gears allows this manipulator to attain a fully integrated light weight design.
Locating actuators at the base of the manipulator, or the use of advanced light weight material and actuator design successfully reduce a link's mass, and thus its associated inertia.
However, this approach in the prior art addresses only half of the problem. Robotic manipulators make use of high performance servo motors to drive their links. The output of these servomotors have inverse characteristics to what is desired when actuating the manipulator. More specifically, servo motors produce low output torque, and at high velocity with respect to what is suitable for most robots. To remedy this, gear reduction systems are most commonly employed. The resulting torque at the link is the actuator torque multiplied by the gear ratio Gr, while the reflected actuator inertia associated with the rotor of the motor is multiplied by Gr2. Thus, the effective inertia experienced by a robotic link can be expressed asJe=Jl+Gr2Jr,  (2)
where Jl is the inertia of the link, and Jr is the rotor inertia of the motor. The reflected actuator inertia of a manipulators can in fact be much larger than that of the link inertia, thereby contributing a larger share of the inertial load during collisions.
In response to this, several prior art actuation systems have been proposed which work to decouple the reflected actuator inertia from the link. However, none of these actuation systems provide enough of a safety margin while providing the desired performance.
Therefore, there is a need for an actuation system which would provide a useful safety margin while providing the required mass and performance.